Identification Of Longjing Tea Varieties Based On Hyperspectral Imaging Technology

The identification of tea varieties depends not only on the appearance(color,shape,etc.)of tea,but also on the internal quality(aroma,taste,etc.).In order to make full use of internal components information and external image information of tea to identify Longjing tea varieties,a method using hyperspectral imaging technology was proposed.The hyperspectral data of six kinds of Longjing tea was captured by a hyperspectral imaging system.First,Principal Components Analysis(PCA)was used to reduce the dimensions of hyperspectral images and select the first two principal component images as feature images.Then,texture analysis based on gray level statistical moment and gray level co-occurrence matrix was implemented on the two feature images.At the same time,RGB model and HSI model were conducted on RGB images where the color features were selected.Next,mean spectral reflectance in region of interest(ROI)from hyperspectral data was acquired.Concluding the spectral analysis methods,the paper chose spectral position variables and developed tea variety indices(TVI)as the spectral features.Finally,Models by using developed BP Neural Network(BP-NN)and Support Vector Machine(SVM)were established to identify Longjing tea varieties,based on image features,spectral features,and the fused features of them respectively.As a result,the right discriminant rate of models based on image features and spectral features were 95%and 98%respectively.But the highest right discriminant rate of model based on the fused features was generally superior to these two models;even the right discriminant rate of 100%can be achieved.This study demonstrated the possibility of applying the hyperspectral imaging technology to more accurately identify the tea varieties.